Is Congestion Pricing Fair? - itf-oecd.org · Is congestion pricing fair? Consumer and citizen perspectives on equity effects Discussion Paper No. 2016-13 Prepared for the Roundtable

13Discussion Paper 2016 • 13

Jonas Eliasson Department for Transport Science, KTH Royal Institute of Technology Stockholm, Sweden

Is Congestion Pricing Fair?Consumer and Citizen Perspectives on Equity Effects

Is congestion pricing fair?

Consumer and citizen perspectives on equity effects

Discussion Paper No. 2016-13

Prepared for the Roundtable on

Income inequality, social inclusion and mobility

(4-5 April 2016; Paris, France)

Jonas Eliasson

Department for Transport Science, KTH Royal Institute of Technology

Stockholm, Sweden

April 2016

The International Transport Forum

The International Transport Forum is an intergovernmental organisation with 57 member countries. It acts as a

think tank for transport policy and organises the Annual Summit of transport ministers. ITF is the only global body

that covers all transport modes. The ITF is politically autonomous and administratively integrated with the OECD.

The ITF works for transport policies that improve peoples’ lives. Our mission is to foster a deeper

understanding of the role of transport in economic growth, environmental sustainability and social inclusion and to

raise the public profile of transport policy.

The ITF organises global dialogue for better transport. We act as a platform for discussion and pre-negotiation

of policy issues across all transport modes. We analyse trends, share knowledge and promote exchange among

transport decision-makers and civil society. The ITF’s Annual Summit is the world’s largest gathering of transport

ministers and the leading global platform for dialogue on transport policy.

The Members of the Forum are: Albania, Armenia, Argentina, Australia, Austria, Azerbaijan, Belarus,

Belgium, Bosnia and Herzegovina, Bulgaria, Canada, Chile, China (People’s Republic of), Croatia, Czech

Republic, Denmark, Estonia, Finland, France, Former Yugoslav Republic of Macedonia, Georgia, Germany,

Greece, Hungary, Iceland, India, Ireland, Israel, Italy, Japan, Korea, Latvia, Liechtenstein, Lithuania, Luxembourg,

Malta, Mexico, Republic of Moldova, Montenegro, Morocco, the Netherlands, New Zealand, Norway, Poland,

Portugal, Romania, Russian Federation, Serbia, Slovak Republic, Slovenia, Spain, Sweden, Switzerland, Turkey,

Ukraine, the United Kingdom and the United States.

International Transport Forum

2 rue André Pascal

F-75775 Paris Cedex 16

[email protected]

www.itf-oecd.org

ITF Discussion Papers

ITF Discussion Papers make economic research, commissioned or carried out in-house at ITF, available to

researchers and practitioners. They describe preliminary results or research in progress by the author(s) and are

published to stimulate discussion on a broad range of issues on which the ITF works. Any findings, interpretations

and conclusions expressed herein are those of the authors and do not necessarily reflect the views of the

International Transport Forum or the OECD. Neither the OECD, ITF nor the authors guarantee the accuracy of any

data or other information contained in this publication and accept no responsibility whatsoever for any consequence

of their use. This document and any map included herein are without prejudice to the status of or sovereignty over

any territory, to the delimitation of international frontiers and boundaries and to the name of any territory, city or

area. Comments on Discussion Papers are welcome.

http://www.itf-oecd.org/

Jonas Eliasson – Is congestion pricing fair?

ITF Discussion Paper 2016-13 — © OECD/ITF 2016 3

Acknowledgements

The data used in this study used comes from a survey developed and carried out in Stockholm, Lyon

and Helsinki by Carl Hamilton, Jonas Eliasson, Karin Brundell-Freij, Kati Kiskilää, Charles Raux,

Stephanie Souche and Juha Tervonen, funded by the ERA-NET programme SURPRICE. The

Gothenburg data collection was carried out by Maria Börjesson, Jonas Eliasson and Carl Hamilton.


4 ITF Discussion Paper 2016-13 — © OECD/ITF 2016

Abstract

This paper discusses and analyses whether congestion charges can be considered to be “fair” in

different senses to the word. Two different perspectives are distinguished: the consumer perspective

and the citizen perspective. The consumer perspective is the traditional one in equity analyses, and

includes changes in travel costs, travel times and so on. Using data from four European cities, the

analysis shows that high-income groups pay more than low-income groups, but low-income groups

pay a higher share of their income. This paper argues that which distributional measure is most

appropriate depends on the purpose(s) of the charging system. The citizen perspective is about

individuals’ view of social issues such as equity, procedural fairness and environmental issues. This

paper argues that an individual can be viewed as a “winner” from a citizen perspective if a reform

(such as congestion pricing) is aligned with her views of what is socially desirable. Using the same

data set, this paper analyses to what extent different income groups “win” or “lose” from a citizen

perspective – i.e., to what extent congestion pricing is aligned with the societal preferences of high-

and low-income groups.



Table of Contents

Introduction ...................................................................................................................................... 7

Background and data ....................................................................................................................... 8

Consumer perspectives on the fairness of congestion pricing ...................................................... 9

Incidence of toll payments across income groups ........................................................................ 10 Broadening the perspective: Incidence of compound self-interest ............................................... 15

Citizen perspectives on the fairness of congestion pricing .......................................................... 21

Opinions about fairness and related political issues ..................................................................... 21 Citizen perspectives on congestion pricing ................................................................................... 24

Conclusions ..................................................................................................................................... 28

Notes ................................................................................................................................................ 30

References ....................................................................................................................................... 31

Appendix: Estimation Results ....................................................................................................... 33

Figures

Figure 1. Share of population who pay various amounts in tolls. .................................................... 10

Figure 2. Average toll payments per income class. .......................................................................... 11

Figure 3. Average toll payments as share of income, per income class. .......................................... 11

Figure 4. Average toll payments per income group, with company car exemption (black) and

without (red). ............................................................................................................................ 14

Figure 5. Support for congestion pricing with respect to toll payments (€/month) .......................... 15

Figure 6. Support for congestion pricing in different income groups, split by those who pay

little or nothing (black) and those who pay moderate or large amounts (blue). ....................... 17

Figure 7. Average compound self-interest per income group, relative to the mean in each city ..... 20

Figure 8. Support for congestion charges across income groups in different cities (two different

years in Gothenburg). ............................................................................................................... 25

Figure 9. Consumer and citizen components of the “utility” of congestion pricing, separated by

income group ............................................................................................................................ 27

Figure 10. Consumer and citizen components of the “utility” of congestion pricing, with “citizen

utility” separated into its subcomponents, and separated by income group ............................. 28



Tables

Table 1. Description of the surveys. ................................................................................................... 9

Table 2. Suits index (overall regressivity/progressivity) of the congestion charges ........................ 12

Table 3. Share of each income class who would pay more than 40€/month in congestion charges 13

Table 4. Estimation results: impact of self-interest variables on attitude to congestion pricing. ..... 18

Table 5. Average agreement (from -3 to 3) with statements. ........................................................... 22

Table 6. Correlation between income and agreement with the statement. ....................................... 23

Table 7. Variables affecting voting response (ordered logit) ........................................................... 25



Introduction

Most transport economists and urban planners would agree that scarce road capacity should be

priced, and would hence support congestion pricing as a way to decrease traffic jams and use scarce

urban land more efficiently. There is also substantial evidence from several cities that congestion pricing

indeed works as intended, and that the aggregate social benefits can by far exceed investment and

operating costs, provided that the system is well designed.

However, perhaps the most pervasive argument against congestion pricing is that it is unfair – a

statement that can be interpreted in several different ways. The purpose of this paper is to discuss and

analyse to what extent congestion pricing is “fair”, in several different senses of the word. The

quantitative analyses use survey data from four European cities: Stockholm and Gothenburg (Sweden),

Helsinki (Finland) and Lyon (France). Stockholm and Gothenburg have operational congestion charging

systems, whereas Helsinki and Lyon do not. In the survey, respondents answered a range of questions

regarding their travel behaviour, their views of fairness and several societal/political questions, and how

they would vote in a hypothetical referendum about congestion pricing.

The purpose is to explore the fairness of congestion pricing from two perspectives, which can be

called the consumer and citizen perspectives1 (Nyborg, 2000; Sagoff, 1988). The consumer perspective

includes how an individual is affected personally: how much tolls she pays, how much travel time she

saves, her valuation of travel time and (if specified) the benefit of the recycled revenues. The citizen

perspective is about what the individual sees as “fair”, “just” or “desirable” from a social perspective,

disregarding her own self-interest. Clearly, these two perspectives are affected by each other. What an

individual thinks is “fair” is often correlated with what will benefit herself – after all, (all) humans are

not saints, at least not on the subconscious level. But just as clearly, opinions about societal issues are not

only determined by self-interest. There is abundant evidence that people’s votes and behaviour are also

affected by other concerns than self-interest, for example concerns about equity, environment and

procedural fairness.

Hence, congestion pricing may be seen as “unfair” in two senses, or both. First, they may be seen as

unfair in a “consumer” perspective, if they hurt low income groups disproportionately: if the poor pay

more in tolls than the rich, if they value their times savings less, or if they get less benefit from the

revenues. Effects can either be measured in absolute terms or proportional to income; I will argue that

which is the most appropriate measure depends on to what extent the charges are (also) a fiscal policy.

The consumer perspective – tolls paid, time gained and revenue recycling – is the traditional perspective

on fairness in equity analyses of congestion charges, and there is an abundant literature (e.g. Eliasson &

Mattsson, 2006; Karlström & Franklin, 2009; Levinson, 2010; Small, 1992). This perspective is analysed

and discussed in the first part of the paper (section 3). Comparing results from the four cities show some

striking similarities, despite different system designs and travel patterns.

Second, congestion charges may be seen as unfair from a “citizen” perspective. This would be the

case if the support (or acceptability) of the fundamental underlying rationality or justification of

congestion pricing differs across socioeconomic groups. For example, imagine a scarce resource which

can be allocated through three alternative mechanisms: pricing, queueing or by some

administrative/bureaucratic decision. Different individuals obviously prefer different mechanisms, for a

variety of reasons, and the same individual may prefer different mechanisms in different contexts. Say



that an individual can be labelled a “winner”, from a citizen point of view, when her preferred allocation

mechanism is the one that is used. Similarly, citizens can be labelled “winners” when societal decisions

regarding, say, environmental regulations or tax progressivity are made in consistency with their

preferences as citizens (which may or may not be aligned with their “consumer” interests). The question

is now whether the share of “winners” on congestion pricing is different across (socio-)economic groups.

This would be the case if congestion pricing is an “elite” project, which is more consistent with richer

and/or more educated groups’ views of what is fair, just or socially desirable. It is known from previous

research that, ceteris paribus, support for congestion pricing is higher among individuals who rate

environmental issues as important, and who perceive pricing as a fair allocation instrument. It is easy to

imagine that high-income groups may view pricing as a fairer allocation mechanism than, say,

administrative decisions – perhaps due to education, or self-interest, or social norms. It is also

conceivable that high-income groups may place a relatively higher weight on environmental benefits.

Whatever the reason, if this is the case, it would be reasonable to conclude that rich groups are “winners”

from a citizen perspective, whether or not they are winners from a consumer perspective. These

questions are discussed and analysed in the second part of the paper (section 4). Comparing results from

the four cities again show striking similarities, despite different political cultures in general and framing

of the congestion pricing issue in particular.

Background and data

The data in this study comes from a survey first designed by a Swedish-French-Finnish team of

researchers, and carried out in Stockholm, Lyon and Helsinki in 2011 (Hamilton, Eliasson, Brundell-

Freij, Raux, & Souche, 2014). Later, two waves of the survey (with some minor modifications) were

carried out in Gothenburg in 2012 and 2013, i.e. both right before and almost one year after Gothenburg

introduced its congestion pricing system (in January 2013) (Börjesson, Eliasson, & Hamilton, 2016). The

references provide more information about the data and its collection. The survey was presented as a

general survey about several transport-related issues; to avoid policy bias, it was deliberately not

presented as a survey specifically about congestion charges.

All the four cities are medium-sized cities with fairly typical European structures and transport

systems. All have a historical city centre encircled by more recently populated areas. Around 80% of

households have access to at least one car. Trips have a predominantly radial pattern, with the main flow

of commuters moving inward in the morning and outward in the evening. Public transport shares vary,

but are much higher than e.g. typical US levels in all the four cities. Transit fares are subsidized around

50%. Stockholm and Gothenburg have operational congestion charging systems, whereas Helsinki and

Lyon do not.

In the survey, respondents were asked how they would vote in a (hypothetical) referendum about

congestion charges. Respondents were presented with different systems in the four cities. In Stockholm

and Gothenburg, the question referred to the actual systems. The Stockholm system was introduced in

2006, and consists of a cordon around the inner city where drivers pay €1 to €2 per passage (both

directions) during weekdays, depending on time of day between 06.30 and 18.30. (The Stockholm

experiences are further described in e.g. Eliasson (2008) and Börjesson et. al. (2012).) The Gothenburg

system, introduced in 2013, consists of a cordon with three additional charging borders located as rays

out from the cordon. Drivers pay 0.8€ – 1.8€ per passage (in both directions) depending on the time of



day, weekdays 06:00-18:30. (Traffic effects are described in Börjesson and Kristoffersson (2015), and

public attitudes in Börjesson, Eliasson and Hamilton (2016).)

In Helsinki, the question referred to a proposed system intensively debated at the time of the survey.

The system was supposed to be based on GPS units in all vehicles, with different tariffs per kilometre in

two zones – an inner zone covering the central area in Helsinki, and an outer zone covering most of

Helsinki. Political support for congestion pricing was never widespread, and at the time of the survey, it

became clear that there was a decisive majority against its implementation. At present, there are no plans

for implementing congestion pricing in Helsinki.

In Lyon, the question referred to a hypothetical system where all cars entering the urban centre

would pay 3€ per day, independent of time of day or day of the week, with a maximum of 50€ per

month.

Table 1. Description of the surveys.

Stockholm Helsinki Lyon

Gothenburg,

2012

Gothenburg,

2013

Date Spring 2011 Spring 2011 Spring 2011 December 2012 December 2013

Method Postal Postal Telephone Postal Postal

Number of

responses

1837 1178 1500 1582 1426

Response rate 43% 39% 37% 40% 38%

The subsequent analyses are based on approximate monthly toll payments, calculated from

respondents’ own statements. In Stockholm, Lyon and Gothenburg, respondents were asked how often

they paid the congestion charge (Stockholm and Gothenburg 2013) or would pay if they drove as today

(Lyon and Gothenburg 2012), and this was converted into approximate monthly payments using data on

the average payment per day. In the case of Helsinki, respondents were asked how many kilometres they

drove in each zone on an average day, which was then also converted into an approximate monthly

payment. Obviously, relying on respondents’ own estimates of their toll payments introduces some

uncertainty, so numerical results need to be treated with some caution. However, the general patterns in

the results are robust enough that this is not a significant problem for the purposes of this paper.

Consumer perspectives on the fairness of congestion pricing

The standard economic welfare effects of congestion charges are made up of four parts: paid tolls,

adaptation costs (adjusting one’s travel pattern to the charges), the value of travel time gains and finally

the benefit of the recycled revenues. In the first section, I will concentrate only on the equity effects of

the paid tolls. This is admittedly only a partial analysis, but nonetheless a relevant one. Regarding

adaptation costs, they generally make up a small part of the total welfare effect2. The value of travel time

gains may be substantial, but unfortunately they are difficult to quantify based on our survey data;

however, an attempt is made in the next subsection. Fortunately, previous research has shown that the

amount of tolls paid is a good proxy for the total welfare effect, including adaptation costs and travel

time benefits (Eliasson & Levander, 2006; Eliasson & Mattsson, 2006). The use of revenues, finally, is



of absolutely crucial importance for the equity effects of congestion charging reform seen as a whole.

However, how revenues are used can be seen as a separate issue, and it is illustrative to analyse the direct

equity effects of congestion charges separately from the equity effects of the revenue use.

Hence, the first subsection analyses the incidence of toll payments across income groups in the four

cities. In the second subsection, more variables relating to self-interest are introduced: the value of travel

time savings, the number of car trips and the number of cars in the household.

Incidence of toll payments across income groups

Distributional profile of toll payments

The four systems differ considerably in their design, and in particular with respect to how much

people [would] pay in tolls on average. Figure 1 shows how many who [would] pay different amounts in

tolls, as a share of the region’s population. In Stockholm, the share who pay high amounts in tolls is low,

whereas in the suggested Helsinki and Lyon systems, large shares of the population would pay rather

substantial amounts. The results from the Gothenburg system are interesting compared to Stockholm:

although the charge per passage is lower than in Stockholm, the Gothenburg system affects a much larger

share of the population, which makes the share of the population who pay high amounts much larger than

in Stockholm.

Figure 1. Share of population who pay various amounts in tolls.

Figure 2 shows average toll payments per income group in the four cities. The left pane shows

results in absolute numbers, while the right pane shows the results normalised by the average toll

payment in each city, to facilitate comparison between cities.



Figure 2. Average toll payments per income class.

Absolute numbers (left) and relative to average toll payment (right).

In all cities, high income groups pay much more than low income groups. Looking at the right pane,

the differences across income groups are surprisingly similar in the four cities, despite the differences in

system design and socioeconomic geography. In Gothenburg and Helsinki, however, the highest income

group pay less than the middle groups. In Helsinki it is because the highest income group tend to live

and work more centrally, and hence drive shorter distances on average. In Gothenburg, it is because

company cars are exempt from congestion charges (according to Swedish tax law), and high income

groups have access to company cars to a much larger extent. The company car exemption is discussed

further below.

However, even if the poor pay less than the rich, they actually pay more relative to their income, as

shown in Figure 3. The left pane shows average toll payments as a share of income3 for each income

class, while the right pane shows the same thing but normalised to allow comparisons across cities.

Figure 3. Average toll payments as share of income, per income class.

Percentages (left), relative to average percentage (right).



The figures reveal that the congestion charges are in fact regressive in all the cities – total payments

relative to income falls with increasing income. The diagrams indicate that the regressivity seems to be

largest in Lyon and smallest in Stockholm. A common measure of the overall regressivity of a tax

instrument is the Suits index (Suits, 1977). A flat-rate tax has Suits index 0, a regressive tax has a

negative Suits index and a progressive tax a positive index. The index is bounded between -1 and 1.

Table 2 shows the Suits indices for the four systems.

Table 2. Suits index (overall regressivity/progressivity) of the congestion charges

City Suits index

Stockholm -0.09

Helsinki -0.09

Lyon -0.16

Gothenburg -0.13

That congestion pricing is regressive in this sense is actually an expected result: a consumption tax

will be regressive if the consumption elasticity with respect to income is lower than 1. Even if driving

(especially in the urban centres) increases with income, it usually increases less than proportionally to

income, which means that most taxes on car driving will be (at least slightly) regressive. The Suits

indices in Table 2 reveal that the analysed systems in Stockholm and Helsinki are slightly regressive,

while the Lyon and Gothenburg systems are moderately regressive. For comparison, Metcalf (1996)

calculates the Suits index of the US sales tax to -0.11; CPPP (2007) calculates Suits indices of the gas tax

and sales tax in Texas to -0.25 and -0.18, respectively; West (2004) calculate Suits indices of a US VMT

tax and a size-differentiated vehicle tax to 0.14 and -0.30, respectively; Eliasson et al. (2016) calculate

Suits indices of the Swedish fuel tax and a differentiated vehicle tax to 0.03 and ¬ 0.09, respectively.

Are the distributional profiles fair?

What, then, is the most appropriate definition of “fairness”? Is congestion pricing fair as long as the

rich pay more than the rich? Or is it fair only when the poor pay a lower share of their income? This is a

question without a clear answer, but a few things can be pointed out.

First, prices are usually the same for everyone, regardless of income or wealth (with the exception

of a few deliberate exceptions such as subsidized healthcare and social housing). Prices of gasoline, cars,

food, clothes, housing and so on do not vary with income4. The social desire for increased equality is

instead usually handled by taxation and welfare systems. The fundamental reasons for redistributing

income rather than letting prices depend on income are two: first, the government can then leave to each

individual to choose how she wants to allocate her income on various goods, according to her own

preferences; second, the price of each good will reflect its “value” in terms of scarcity and/or production

cost, so having the same price for everyone will achieve a Pareto efficient outcome. Now, the purpose of

a congestion charge is to correct the price of car driving for external effects, to make it better reflect the

total social cost of car driving. In other words, the price for car driving with the congestion charge is

what the price really should be; without the charge, driving is subsidised from a social point of view.

From this perspective, it can in fact be argued that the distributional effects of introducing congestion

pricing are irrelevant – that is, if one accepts that the default situation is that prices are equal for



everyone, and should reflect the true, social cost for each good. At least, one should realize that arguing

against corrective taxes with distributional arguments is logically equivalent to arguing that the good in

question (car travel, for example) should be subsidized for distributional reasons – and this is often a

much less persuasive or intuitively appealing argument.

Second, however, congestion charges may well have a strong fiscal motivation as well. The

Gothenburg system is a good illustration: the system was designed to generate a revenue of at least 1

billion kronor (100 M€) per year to be used for infrastructure investments. As a secondary purpose, the

system should decrease congestion as efficiently as possible, given this revenue constraint. In such cases,

congestion pricing is clearly not only about correcting prices; the purpose is at least as much to generate

public revenues. This makes it more appropriate to compare distributional effects of charges against

income taxation; had the revenues not come from congestion charges, they would have had to come from

the usual public tax sources. Hence, comparing toll payments relative to income is a natural default

position in such cases.

Third, being aware of the distributional effects of any new policy is clearly important. Any change

in prices causes transition costs which may be important to consider, at least for determining at which

speed a change can be implemented. Moreover, real congestion pricing systems are not perfect – they do

not, in reality, perfectly reflect the true social cost of each car trip due to technical or cognitive

constraints; some car trips will actually be overpriced, while others will still be underpriced. In this

perspective, it can be relevant to check how many drivers experience a substantial increase in travel

costs, defining “substantial” in some suitable way. If this share is high, especially in low-income groups,

it may be a warning signal indicating either that the policy may be too ambitious too fast, or that the

design punishes some trips disproportionately.

Looking only at averages hides the fact that the variation within each income group may be

substantial. Compare this to an income tax, which will by definition affect everyone with the same

income in the same way. In the case of congestion charges, there may be subgroups who are hurt

disproportionately relative to their income, even if the charge is progressive overall (or at least not very

regressive). This is in fact often the most important argument of those arguing against congestion charges

using distributional arguments: not that the policy is necessarily very regressive on average, but that

there may be non-negligible subgroups who are hurt disproportionately.

With this in mind, Table 3 shows the share of each income group who [would] pay more than 40

€/month in congestion charges. In Stockholm and Gothenburg, the shares are (very) small in the lower

income groups, although they may still be non-negligible of course. In the suggested Lyon and Helsinki

systems, however, the shares are quite high even in the lowest income groups, meaning that there are

quite a few people who would see their driving costs rise considerably even among the poor. Remember,

though, that these systems were never properly evaluated and re-designed. The Helsinki system was a

real suggestion, but never made it further than the initial political debate. The Lyon system was designed

for the purpose of the survey. Hence, it seems that real congestion pricing systems can be designed to

have a much smaller impact on travel costs.

Table 3. Share of each income class who [would] pay more than 40€/month in congestion charges

€/month 1000 2000 3000 4000 5000

Stockholm 2% 7% 7% 13% 15%

Helsinki 15% 26% 35% 53% 47%

Lyon 18% 31% 36% 37% 48%

Gothenburg 7% 18% 29% 32% 22%



Lessons from the Swedish company car exemption

The Swedish company car exemption is an example of a generally important point, namely that the

distributional effects of congestion charges depends on the specific design of the charging system, and

that legal decisions may have unintended consequences. Some time after congestion charges had been

introduced in Stockholm, the tax court determined that congestion charges were to be considered part of

the operating costs of the car, and as such they were included in the “taxable benefit value” of a company

car. This meant that company car owners either paid no charge at all, or could deduct the charge from

their before-tax salary (which implied a substantial discount), depending on the company’s policy. This

was a completely unintended consequence of how the tax law interacted with the legal definition of the

congestion charge (which, in legal terms, is a national tax) – but it had substantial effects on the

distributional profile of the charges, especially in Gothenburg with its high prevalence of company cars

in high income groups (the city is dominated by the car industry).

Figure 4 shows the effect of the exemption. Without it, the highest income group had paid the most;

with it, the richest group pay on average as little as the second-lowest income group. Further, the

regressivity had been much smaller without the exemption: the Suits index had been 0.06 rather than

0.13. The effect in Stockholm seems to be smaller, but unfortunately, there is so far no data available for

Stockholm to analyse this in depth.

Figure 4. Average toll payments per income group, with company car exemption (black) and

without (red).

Absolute numbers (left) and as share of income (right).

Tax rules for company cars are complicated in many countries, and the design of these rules may

have profound and substantial effects on travel patterns and equity (see for example an analysis of a

change in the UK tax rules by Le Vine et al. (2013)). As the Gothenburg example shows, it may

drastically change the distributional profile of congestion pricing. More generally, the lesson is that

exemptions of various kinds – for residents, professional traffic, certain residential areas and so on – may

have important and perhaps unintended consequences both for the effectiveness and the distributional

profile of congestion pricing.



Distributional effects in other dimensions

In addition to income, congestion pricing can have distributional effects across other socioeconomic

characteristics as well. Perhaps surprisingly, these differences are relatively small once income is

controlled for. Detailed results are omitted to save space. The main differences across socioeconomic

groups (controlling for income) are:

In Lyon and Gothenburg, households with children under 18 years of age [would] pay more

tolls (controlling for income) than households without children.

In all cities except Gothenburg, men [would] pay more than women (even controlling for

income differences). The difference is largest, in relative terms, for middle income groups.

In Lyon, people older than 65 years would pay less (after controlling for income). In the other

cities, differences across age groups are negligible.

Education does not affect average toll payments in any systematic way.

Broadening the perspective: Incidence of compound self-interest

How much someone pays [would pay] affects the person’s attitude to a suggested congestion

pricing system: all else equal, people are more negative the more they [would] pay (e.g. Börjesson et al.,

2016; Eliasson, 2014; Eliasson & Jonsson, 2011; Gaunt, Rye, & Allen, 2007; Hamilton et al., 2014;

Hårsman & Quigley, 2010; Jaensirisak, May, & Wardman, 2003; Schade & Schlag, 2003). Respondents

were asked how they would vote in a referendum about congestion pricing, regarding the actual schemes

in Stockholm and Gothenburg, the debated scheme in Helsinki, and a hypothetical area charging scheme

in Lyon. The response alternatives were Certainly yes, Probably yes, Probably no, Certainly no or No

opinion/I don’t know. Answers clearly correlated with the amount of tolls respondents paid or would

pay. Defining “support” as the share of positive responses excluding “No opinion/I don’t know”,

Figure 5 illustrates how support depends on the [anticipated] toll payments.

Figure 5. Support for congestion pricing with respect to toll payments (€/month)



Clearly, support falls as toll payments increase. Note, however, that the biggest drop in support is

between those who pay no toll at all (e.g. don’t own a car) and those who pay something, albeit just a

little (Lyon is an exception). It is interesting how similar the patterns are in the three cities without

congestion charges: Helsinki, Lyon and Gothenburg in 2012 (before the system started). The figure also

illustrates the “experience” effect in Gothenburg: the Gothenburg 2013 curve has moved upwards with

almost precisely the same shift for all groups, regardless of toll payments (although the effect is slightly

smaller for those who pay no toll at all).

In addition to toll payments, attitudes are also affected (as we shall see) by several other variables

which are related to self-interest, and should hence be included in the consumer perspective: how many

car trips a respondent makes, her value of travel time savings and how many cars there are in the

household. That the number of cars and car trips significantly affect attitudes to congestion pricing even

after controlling for toll payments may be for several reasons: it may reflect the general car dependency

of the household or individual, for example.

The fact that there are more variables than just toll payments which affect attitudes to congestion

pricing means that an analysis which only takes toll payments into account may not fully reflect the

subjective benefits and losses of charges, as perceived by the individual. If the aim is to measure the

incidence of congestion pricing as perceived by the individuals themselves, the analysis has to be

extended to account for these other variables as well. After all, these variables also reflect various aspects

of the perceived personal incidence of the congestion charges. I will call this total, perceived, personal

incidence the compound self-interest, since it is the sum of several factors: tolls paid, time gains,

adaptation costs and so on.

How, then, should these variables be weighted together? A natural approach is to estimate a

statistical model of how they affect respondents’ attitude to congestion charges, and take the estimated

model parameters as relative weights of the different variables. In this way, the variables act as indicators

of or proxies for different aspects in which congestion charges affect the individuals, which in turn

allows for a richer analysis of the distributional effect of congestion pricing.

In addition, it is desirable to take into account that variables may affect different income groups

differently. For example, it might be natural to expect that the same amount of toll payments might cause

more disutility for low income groups than for high income groups, simply because low income groups

might have higher marginal utility of income. Exploring the data, however, gives only limited support for

this hypothesis. Figure 6 shows that there are only a few cases where it seems as if (high) toll payments

affect attitudes differently across income groups.



Figure 6. Support for congestion pricing in different income groups, split by those who pay little or

nothing (black) and those who pay moderate or large amounts (blue).

Cross-tabulations of this kind only give rough indications, of course. To properly separate the

impact of different variables, respondents’ attitude to congestion pricing (how they would vote in a

referendum) is regressed on various self-interested-related variables. A potential problem with the model

estimation is that attitudes are affected by several other variables as well – in particular other attitudes –

and if we omit these, the parameter estimates may be biased if there are correlations between included

and omitted variables. Fortunately, however, this turns out not to be a problem: when attitude variables

are introduced in the model (see next section), this does not change the parameters for the self-interest

variables appreciably. Moreover, the general conclusions are robust for various other model

specifications.

Since the voting response is an ordered variable, an ordered logit model is used. A comprehensive

description of ordered models can be found in Greene (2003), but a short intuitive description is



necessary. Let y be a latent (unobservable) variable, which is a linear function of a vector of observable

variables X, a parameter vector β (to be estimated) and an idiosyncratic term ε

𝑦 = 𝛽𝑋 + 𝜀

The latent variable y can be interpreted as a measure of how positive an individual is to congestion

charges, which cannot be observed (measured) directly. What is observable is the voting response z,

which has five ordered levels z=1,…,5 from most negative to most positive. We assume that this

response is determined by in which of five intervals y falls. The limits of the intervals are determined by

estimated threshold parameters μ_k, so we have

𝑧 = 1 if 𝑦 ≤ 𝜇1

𝑧 = 2 if 𝜇1 ≤ 𝑦 ≤ 𝜇2

(…)

𝑧 = 5 if 𝜇4 ≤ 𝑦

Assuming that ε is logistically distributed, the probabilities that z=i become

𝑃1 = 1 −1

1 + exp(𝜇1 − 𝛽𝑋)

𝑃𝑖 =1

1 + exp(𝜇𝑖 − 𝛽𝑋) −

1

1 + exp(𝜇𝑖−1 − 𝛽𝑋) 𝑖 ∈ {2,3,4}

𝑃5 =1

1 + exp(𝜇4 − 𝛽𝑋)

Table 4 shows the estimation results from the ordered logit model, indicating how respondents’

attitude to congestion charges (as measured by the voting response) is affected by various self-interested-

related variables. A large number of model specifications have been tested, but details are omitted to save

space. Positive parameters indicate that support increases when the variable increases.

Table 4. Estimation results: impact of self-interest variables on attitude to congestion pricing.

Value Std. Error t value Tolls -0.0010832 0.0001125 -9.6249 Tolls, add. inc.grp 1 -0.0007957 0.0002209 -3.6030 No car 0.3061705 0.0610495 5.0151 Car trips 0.4306097 0.0284951 15.1117 Value of time, drivers 0.2389846 0.0189181 12.6326 Stockholm 0.9948645 0.0668150 14.8898 Helsinki 0.0213867 0.0740519 0.2888 Lyon -0.1766564 0.0722816 -2.4440 Gothenburg2014 0.5572110 0.0697637 7.9871 Intercepts: Value Std. Error t value 1|2 0.4265 0.1022 4.1721 2|3 1.4528 0.1034 14.0502 3|4 1.9839 0.1047 18.9533 4|5 3.3649 0.1098 30.6583 Residual Deviance: 19960.01 AIC: 19986.01

Toll payments affect attitudes negatively; the effect is proportional to the amount of tolls paid. This

effect is stronger for the lowest income group (<1500 €/month) (the two “tolls” parameters are added),

but there are in fact no significant differences among the other income groups. Not owning a car at all



increases support more than can be explained merely by not having to pay tolls, so a dummy variable for

“no car in household” is significant and positive. More car trips decrease support (with a linear effect),

even after controlling for toll payments. The model also includes dummy variables for the different cities

and for the difference in Gothenburg between 2012 and 2013; Gothenburg 2012 is taken as the reference

level. The attitude differences between Gothenburg 2012, Lyon and Helsinki (neither of which had

congestion pricing in place) are very small, while attitudes in Stockholm and Gothenburg 2013 (after

congestion pricing was introduced) are much more positive.

Finally, the value of travel time savings has a substantial positive impact on support for those who

make at least a few car trips per week. The value of travel time savings was measured with the following

thought experiment:

You commute daily by car. On the way, you have to cross a bridge5 across a river. One day, the

bridge closes for repairs for a long time. Another bridge is available further downstream, but the detour

takes an additional 20 minutes. During the time it takes to repair the bridge, the road authority has

arranged with a ferry that can take cars over the river. What is the highest amount you would be prepared

to pay for a one-way ticket for the ferry, to save 20 minutes on your journey to work?

Respondents were given seven alternative answers ranging from “nothing” to “more than 5€”.

For our purposes, it is important to note that the impacts of the self-interest-related variables do not

vary systematically across income groups (except for toll payments in the lowest income group).

Surprisingly, the effect on attitudes of making one car trip more or paying one more euro in tolls and so

on seems to be the same, regardless of income. A plausible hypothesis is that for the moderate monetary

amounts we are considering, the differences in marginal utility of money are not large enough to matter

for attitudes.

With the model results in hand, we can calculate a broader measure of the perceived incidence of

congestion charges across income groups, simply by calculating the relative differences in the latent

variable y across income groups. Note that this is an exploratory analysis where the numerical results

must be interpreted with caution: there is no guarantee that y can be interpreted cardinally, or in other

words, that the absolute magnitude of y is meaningful (since the “unit” of y may not be constant if the

distances between thresholds are very different). However, comparing average values of y across income

groups will give an indication whether some income groups can be said to be worse off6.

Figure 7 shows the distributional profile across income groups for the four cities. Results have been

normalized for each city, since the purpose is just to compare the relative tendency across income groups

for each city. (Let 𝑦𝑛 be the latent variable for individual n. The bar in the histogram corresponding to

income group i in city j is 𝐵𝑗𝑖 =

∑ 𝑦𝑛𝑛∈(𝑖,𝑗)

∑ 𝑦𝑛𝑛∈𝑗− 1. In other words, the diagram shows how much better or

worse off each income group of a city is, on average, compared to the average citizen in that city.

(Remember that this measures only “self-interest” effects, weighted as individuals perceive the effects

themselves, as measured by their voting response.)



Figure 7. Average compound self-interest per income group, relative to the mean in each city

Except for Lyon, the pattern is the same in all cities: higher income groups are worse off, but the

highest income group is in fact a little better off than the second-highest. This is partly due to this group

having high values of time (on average), and partly due to average toll payments being lower than for the

second-highest group due to central residential locations (Helsinki) or company car exemptions

(Stockholm and Gothenburg). The relative differences across income groups are smaller in Stockholm

than in Gothenburg and Helsinki. For Lyon, the distributional profile of the suggested system is different:

the lowest income group is better off than the average, but after that, higher income groups are better off

than lower income groups. The primary reason for this seems to be the design of the (hypothetical) Lyon

system, where all car trips in the entire urban area is charged the same amount, regardless of time of day,

destination or distance.

Broadly speaking, this shows that congestion pricing is “progressive” (with a slight abuse of the

term) in the sense that lower income groups are hurt less by direct self-interest effects – as perceived by

the individuals themselves, as measured by how self-interest variables influence voting response. The

exception is Lyon, but the reason for this seems to be the coarse design of the system: a more realistic

and efficient design may well have yielded other results. The lesson that can be drawn from Lyon is,

again, that the design of the system is crucial for the distributional profile.



Citizen perspectives on the fairness of congestion pricing

The previous section only dealt with fairness from a consumer perspective: how congestion pricing

affects different income groups in terms of self-interest variables such as money, time and so on. This

section analyses fairness from a citizen perspective. Different allocation mechanisms, and different social

goals or benefits, can be viewed by citizens as “fair” in a more abstract sense. In this perspective, factors

such as procedural justice, equity, equal treatments, human rights and the relative weights of different

social objectives often matter (depending on the issue). Citizens’ perceptions and definitions of fairness

will vary, of course, and may well correlate with their self-interest – although the direction of causality is

not always easy to establish – but this does not change the point that “fairness” is something that people

apparently value in itself, even abstracting from their own self-interest.

Here, I will focus on one particular issue, namely whether the opinions of congestion pricing from

this citizen perspective vary systematically across socioeconomic groups – in particular across income

groups. In other words, is the concept of congestion pricing – the principle of allocating scarce road

space according to willingness to pay – more consistent with, say, high-income groups’ views of what is

“fair” or “just” than with low-income groups’ views? If so, one could reasonably argue that congestion

pricing is an “elite” project, since the concept would be better aligned with what high-income groups

view as a “fair” or “just” society. A priori, I see no particular reason to expect neither this nor the

opposite; but I think the issue clearly matters for the socio-ethical or democratic justification of

congestion pricing, and the fact that I do not know what result to expect a priori makes the question

interesting.

Disentangling this question, however, is complicated. Obviously, simply asking respondents “is

congestion pricing fair?” will elicit responses so tainted with self-interest that it is virtually pointless.

Instead, our survey contained a large number of questions about whether respondents perceived various

allocation mechanisms as “fair”, and also questions about other social issues such as environment and

income equity. Through econometric modelling (controlling for self-interest variables), we can then

reveal how perceptions of these more or less related issues correlate with the attitude to congestion

pricing, and finally measure how well aligned congestion pricing is with these related socio-political

views.

The question of how perceptions of fairness and other social issues vary across income groups and

cities is of course interesting in itself. Therefore, the first subsection explores this, before the second

subsection turns to our main issue of citizen perspectives of congestion pricing.

Opinions about fairness and related political issues

Respondents were presented with a number of statements, and asked to indicate to what extent they

agreed on a 7-grade scale, from “completely disagree” (-3) over “neither agree nor disagree” (0) to

“completely agree” (+3). Table 5 shows how respondents in the four cities agreed, on average, with the

statements.



Table 5. Average agreement (from -3 to 3) with statements.

Stockholm Helsinki Lyon Gothenburg

1 “Considerably more resources should be used

to protect the natural environment.” 1.4 1.3 2.1 1.3

2 “The government should prioritise to reduce

the differences between the poor and the rich

in the society.”

0.9 1.2 1.7 1.3

3 ”Taxes in [country] are too high” 0.8 1.2 1.3 0.3

4 Pricing the ferry is fair 1.9 1.4 0.3 0.9

5 Queueing to the ferry is fair 1.5 2.1 -1.2 0.8

6 Letting a public agency decide about space on

the ferry is fair 0.1 -1.1 -1.6 -0.7

7 Giving out places on the ferry with a lottery is

fair -1.1 -1.3 -2.3 -2.2

8 “It is fair [justified] that airplane tickets cost

more for departure during peak hours than

during off-peak”

0.9 0.8 -0.4 0.3

9 “It is fair [justified] that airplane tickets to

vacation destinations cost more when the

weather in [country]is bad.”

-1.0 -1.2 n/a -1.4

10 “It would be fair [justified] if transit fares

were lower in off-peak hours” 1.0 0.6 0.7 0.9

Statements (1)-(3) are about general political issues: environment, social equity and taxes.

Respondents broadly agree with the statements, on average. Almost no one disagrees with the

environment statement (2). Swedish respondents agree less with the statement that taxes are too high (3).

Comparing Stockholm and Gothenburg, more respondents in Gothenburg agree with the equity question

(2) and disagree with the “taxes are high” question (3).

Statements (4)-(7) are perhaps the most interesting and relevant in the context of this paper.

Following the thought experiment about a ferry replacing the broken bridge (see section 3.2),

respondents were given the following question:

Some people complain that it is unfair that the authority charges a price for the ferry tickets. When

offering the ferry for free, it turns out that there is not room on the ferry for everyone who wants to use it.

The authority now considers four different methods to choose who gets to travel with the ferry:

Price: Revert to the original policy of charging those who want to travel, and set the price so

the ferry is just filled.

Queue: Those who arrive first to the jetty and stand first in line get to go with the ferry.

Authority determines “need”: Those who want to travel with the ferry have to show some

evidence to support their need. The authority then provides ferry passes based on their

judgment of the greatest need.

Lottery: Tickets are allocated randomly, so that everybody has an equal chance of winning.

To what extent do you consider these alternatives fair? [7 grade scale from “Completely unfair” (-3)

to “Completely fair” (+3)]



In almost all cities, respondents rate the fairness similarly: pricing comes first, followed by

queueing and then decisions by a public authority, with lottery as a distant fourth. Helsinki differs in that

queueing is ranked as fairer than pricing. However, answers vary considerably across countries, and are

actually in line with some clichés about national political cultures. French respondents rate all

alternatives lower than respondents from other countries; in fact, 25% of French respondents rate all

allocation mechanisms as more or less unfair (below 0) and 17% rate all allocation mechanisms as “very”

or “completely” unfair (begging the question if there is any “fair” mechanism to allocate space on the

ferry). Swedish respondents seem to put more trust in public authorities (which is consistent with many

other studies), rating decisions by a public agency as fairer than respondents from other countries do.

Statements (8)-(10) concern various forms of scarcity pricing in other contexts. Differentiating air

fares (8) and transit fares (10) with respect to peaks in demand is rated “fair” by a narrow majority of

respondents in most cities – although in all cities, there are sizeable groups who rate this as unfair.

However, commercial airlines taking advantage of increased demand due to bad weather strikes a

majority of respondents as unfair. One interpretation is that there is a difference in perceived fairness of

pricing between situations when supply is “necessarily” scarce – such as airport and transit capacity in

peak hours – and situations where commercial companies simply extracts an increased willingness to

pay.

For the purpose of this paper, the most interesting question is whether the answers vary

systematically with income – in particular, whether views of the fairness of allocation mechanisms

(including pricing) do. Table 6 shows correlations between income and agreeing with the statements.

Table 6. Correlation between income and agreement with the statement

Stockholm Helsinki Lyon Gothenburg

1 “Considerably more resources should be used

to protect the natural environment.” -0.06 -0.10 -0.02 -0.04

2 “The government should prioritise to reduce

the differences between the poor and the rich

in the society.”

-0.19 -0.27 0.00 -0.17

3 ”Taxes in [country] are too high” 0.06 -0.03 -0.01 0.02

4 Pricing the ferry is fair 0.00 0.01 0.01 0.02

5 Queueing to the ferry is fair -0.08 -0.03 0.03 -0.05

6 Letting a public agency decide about space on

the ferry is fair 0.02 0.01 -0.02 0.01

7 Giving out places on the ferry with a lottery is

fair -0.05 -0.02 0.01 -0.02

8 “It is fair7 [justified] that airplane tickets cost

more for departure during peak hours than

during off-peak”

0.09 0.01 0.03 0.13

9 “It is fair [justified] that airplane tickets to

vacation destinations cost more when the

weather in [country]is bad.”

0.09 0.02 n/a 0.11

10 “It would be fair [justified] if transit fares

were lower in off-peak hours” 0.00 -0.06 0.02 -0.03

The most striking result is that the correlation between agreement and income is generally very low.

The main exception is only that low income groups agree much more with the equity question (2) (except

in Lyon). Particularly surprising is the small differences across income groups in how respondents rate



the fairness of pricing mechanisms (4,8,9,10) and whether taxes are too high (3). In Sweden, high-

income groups have a slight tendency to agree more with air pricing questions (8,9) and that taxes are too

high, but the tendency is small. More detailed analyses, also considering the tails and asymmetries of the

distribution of answers, reveal a few more observations:

Rich tend to agree more that having higher airfares in peak hours (8) is fair/justified. In

particular, the share who agree strongly with the statement is higher in rich groups.

Rich tend to agree slightly more that it is fair/justified that airfares to vacation destination are

higher when the weather is bad (9). In particular, the share who strongly disagree is smaller

among rich groups.

In all cities except Helsinki, opinions about taxes (2) follows a U-shaped pattern: the two

lowest income groups agree the most with that taxes are too high, but the highest income group

agree more with this statement than the two middle income groups. In Helsinki, there is no

correlation between income and opinions about taxes.

Slightly fewer among the rich think that more resources should be spent on protecting the

environment. In particular, fewer of the rich agree strongly; almost no one, in any income

group, disagrees with the statement. In Stockholm and Gothenburg this tendency is small,

whereas it is considerable in Lyon and Helsinki (the share who agree strongly (≥2) falls from

80% in the poorest group to 60% in the richest group in Lyon, and from 60% to 40% in

Helsinki).

Comparing which way to allocate space on the ferry is rated as the fairest by each individual

reveals that a higher share of the rich rates pricing highest on the fairness scale. The difference

is not big, however: in Stockholm, where this difference is largest, 50% in the richest group

rate pricing as most fair, whereas 37% in the poorest group rate pricing as most fair.

Conversely, slightly more in the poor groups rate allocating ferry space by a public authority as

the most fair. More of the rich strongly disagree that this is a fair allocation mechanism. (There

are no differences in how queueing is rated, however.)

As already noted, a large share of French respondents rate all alternatives to allocate ferry space

as unfair (in the other countries, this share is negligible – less than 2% of respondents). There is

no difference across income groups in this respect.

Summing up these findings, a higher share of the poor agree with the equity statement (2), and a

slightly higher share among the rich regard pricing as a fair allocation mechanism. On average, slightly

fewer of the rich express strong environmental concerns (1). With these findings in hand, the next section

explores correlation between these attitudes and the support for congestion pricing.

Citizen perspectives on congestion pricing

The main purpose of this section is to explore how “citizen” perspectives of congestion pricing vary

across income groups, as explained above. An obvious place to start is to check whether the support for

congestion pricing varies across income groups. As Figure 8 reveals, this is not the case: there is no

particular systematic tendency in how different income groups would vote about congestion charges.



Figure 8. Support for congestion charges across income groups in different cities (two different years in Gothenburg).

However, these aggregate figures do not reveal enough, since voting responses are affected by a

mixture of self-interest and other considerations. Instead, we estimate a revealing how the attitude to

congestion pricing is affected both by self-interest variables and by a number of variables representing

citizen perspectives. Since several of the attitude questions (Table 5) measure similar things (in particular

the attitude to scarcity pricing in different contexts), a subset of these indicators has to be selected. After

testing various options, the following variables are included in the model: environment (agreement rating

of statement (1) in Table 5), equity (statement 2), taxes are too high (statement 3), pricing is fair (4) and

agency decision is fair (6). All these variables can be expected to correlate with the attitude to congestion

pricing. To what extent they do may depend on how congestion pricing is perceived or framed in the

local public debate or discourse. For example, if congestion charges are perceived as “just another tax”,

then the correlation with the attitude to taxes can be expected to be strong; if congestion charging is

perceived as an environmental policy, then the correlation with the attitude to environmental policy can

be expected to be strong – and so on. Table 7 shows estimation results from the ordered logit model (a

binary logit model distinguishing only positive/negative answers give similar results).

Table 7. Variables affecting voting response (ordered logit)

Value Std. Error t value Tolls -0.0010111 0.0001179 -8.5742 Tolls, add. inc.grp 1 -0.0005993 0.0002281 -2.6274 No car 0.3297529 0.0641394 5.1412 cartrips 0.3327746 0.0302396 11.0046 Value of time, drivers 0.1825060 0.0199794 9.1347 Stockholm 1.0063928 0.0713684 14.1014 Helsinki 0.1555103 0.0780314 1.9929 Lyon -0.0990411 0.0771785 -1.2833 Gothenburg2014 0.5615992 0.0776810 7.2296 environ 0.2609645 0.0166537 15.6701 TaxTooHigh -0.2524060 0.0118604 -21.2813 PricingIsFair 0.1172166 0.0114416 10.2447 AgencyIsFair 0.0504215 0.0098760 5.1054 equity -0.0042233 0.0134644 -0.3137 Intercepts: Value Std. Error t value 1|2 1.0315 0.1610 6.4059



2|3 2.1687 0.1627 13.3295 3|4 2.7289 0.1639 16.6526 4|5 4.2695 0.1684 25.3584 Residual Deviance: 17698.47 AIC: 17734.47

The parameters of the self-interest variables are broadly unchanged by the introduction of attitude

variables. The first four are highly significant: environment and pricing is fair is associated with more

positive attitudes, while taxes are too high is associated with a more negative attitude. Authority decision

is fair is also associated with more positive attitudes, although this effect is smaller. At first this may

seem counterintuitive: after all, approving of authority decisions based on subjective “need” may seem

like the opposite to approving of market-based solutions based on willingness to pay. However, what this

question measures is rather the trust in government – whether the respondent believes that public

agencies are, on average, trustworthy. Previous research has shown that support for congestion pricing

correlates with trust in government and also supporting various kinds of public interventions (e.g. speed

cameras).

Two negative findings in the model estimation are interesting. First, the equity variable is not

significant (in any of the cities): there is, surprisingly, apparently no correlation between respondents’

opinions about equity and their opinions about congestion pricing. In the light of this, the preoccupation

with congestion charges’ equity effects is rather remarkable – unless it is simply because supposedly

negative equity effects is a more convenient argument against congestion charges in public debate than

self-interest arguments. Second, there are no differences in the parameters for the attitude variables

across income groups. In other words, attitudes affect voting responses in the same way, regardless of

income. It might have been natural to expect that, for example, rich groups could “afford” to let, say,

environmental concerns affect their attitude to congestion pricing more than poor groups – but this is

apparently not the case.

The model in Table 7 shows how self-interest variables and the various “citizen perspective”

variables are weighted against each other. This means that we can now interpret the underlying latent

variable as an extended “utility function” comprising two parts: a consumer part (consisting of the self-

interest variables) and a citizen part (consisting of the citizen perspective variables). Note that this is an

exploratory analysis, and the same caveat applies as noted previously: there is in principle no guarantee

that the latent variable can be interpreted cardinally. However, exploring its average over income groups

will give an indication of how attitudes are affected by consumer and citizen aspects, and how this varies

across income groups. The robustness of the conclusions has also been checked with two other methods

(a binary logit model and a predicted-response method; see endnote 5).



Figure 9. Consumer and citizen components of the “utility” of congestion pricing,

separated by income group

Figure 9 illustrates the citizen and consumer components of this “extended utility function” of

congestion pricing. The blue bars (the “self-interest” or “consumer” component) are essentially the same

as in the analysis of compound self-interest (Figure 7) but not rescaled with the average in each city. As

was noted before, the general tendency is that low-income groups fare better compared to the average in

each city in terms of the self-interest component, although the highest income group actually fare better

than the second-to-highest.

The red bars – the “citizen” part – is the new part. Remember that this part of the “utility” function

can be interpreted as how well congestion pricing is aligned with individuals’ opinions about

environmental policy, fairness of pricing as an allocation instrument, the level of taxation and trust in

government. An individual who scores highly on these attitudes will tend to be a “winner” in a citizen

sense (controlling for the self-interest component) if congestion pricing is introduced, in an analogous

way as an individual with high value of time, low toll payments and so on will be a “winner” in a

consumer sense (controlling for the citizen component)

The citizen effects are clearly different across income groups. Except for Helsinki, the general

tendency is that the middle income group “wins” more than the low- and high-income groups. In

Helsinki, income groups simply “win” more the lower income they have.

To explain this in more detail, Figure 10 separates the “citizen utility” into its subcomponents.

Analysing this in detail shows that the reason that middle-income groups are “winners” from a citizen

perspective is primarily because they are the most content with current tax levels, and also rate

environmental concerns highly (on par with the lower income groups but slightly higher than the two

highest income groups). It is also evident that the attitude variables, taken separately, have a very strong

impact on attitudes. Whether congestion pricing is perceived as, for example, mostly an environmental

policy or mostly as a tax will hence have a vast impact on support for congestion charges.



Figure 10. Consumer and citizen components of the “utility” of congestion pricing, with “citizen

utility” separated into its subcomponents, and separated by income group

Is congestion pricing fair from a citizen perspective?

Given these results, can congestion pricing be considered “fair” from a citizen perspective? My

conclusion would be a qualified “yes”. Clearly, congestion pricing is not more aligned with the citizen

preferences of the “elite”, defined as the high(est) income groups. In three of the case cities, it is the

middle income group who seem to “win” the most in a citizen perspectives; put differently, if is in this

group that congestion pricing most closely aligns with the group’s societal preferences. Perhaps the most

important result, however, is that most differences in citizen preferences are small or negligible across

income groups. Consequently the distribution of “citizen utility” is rather similar across groups.

The only social issue where there is a clear difference in opinions across income groups is equity.

Lower income groups agree to a much larger extent that society needs to prioritise to reduce the gap

between rich and poor. However, this turns out not to be related to the congestion pricing issue: there is

no correlation between views of equity in any of the case cities.

Conclusions

The purpose of this paper has been to discuss and analyse to which extent congestion pricing can be

viewed as fair. Since “fairness” can be interpreted in several different ways, a range of different analyses

have been presented.

Starting with the consumer perspective on fairness, the incidence of toll payments is strongly

correlated with income: high-income groups pay more in tolls, on average, in all cities. However, since

car driving tends to increase less than proportionately with income, low-income groups tend to pay more

tolls relative to their income, on average. A Suits index calculation shows that all of the four analysed

congestion charging systems are regressive. Whether this is a problem depends, in my view, on the



purpose of the charges. If the purpose is truly to correct the price of car driving, to make it reflect the full

social cost of driving, then it is in fact doubtful whether distributional effects are relevant, at least in the

long run. Amending income inequalities through taxation and welfare systems is both more effective and

efficient than subsidising goods; and allowing prices to be lower than the social cost is equivalent to a

social subsidy. However, congestion charges often have dual purposes: in addition to being a corrective

tax, the purpose is often also (and sometimes primarily) to generate revenues, often for infrastructure

investments. To the extent that the charges is a fiscal measure, the potential regressivity of congestion

charges is a serious problem. After all, it is difficult to see why the poor should contribute more than

proportionately to public tax revenues.

The distributional profile of congestion pricing depends on the design of the specific system –

location of charging points or areas, exemptions, time of day and so on. An illustrative example is the

Swedish (unintended) exemption for company cars, which turned the Gothenburg system from an almost

equity-neutral to a clearly regressive system.

In a second analysis, the consumer perspective was broadened from merely toll payments to a range

of variables relating to self-interest: access to car in the household, value of time savings and the number

of car trips. The weights of the different variables were obtained by estimating their impact on

individuals’ attitudes to congestion charges (how they would vote in a congestion pricing referendum).

One interesting finding is that income hardly matters for how these variables affect attitudes to

congestion charges; their impact on attitude is almost the same across all income groups (except the

lowest group). With these weights, a compound self-interest measure was calculated, and its incidence

across income groups was explored. Broadly speaking, this analysis showed that congestion pricing is

“progressive” (with a slight abuse of the term) in the sense that lower income groups are hurt less than

average by the direct, self-interest effects, as perceived by the individuals themselves. The exception is

Lyon, but the reason for this seems to be the coarse design of the system: a more realistic and efficient

design may well have yielded other results.

Fairness can also be viewed in a citizen perspective. Depending on individuals’ views of procedural

fairness, equity, environmental issues and so on, congestion pricing can be viewed as more or less “fair”

in an abstract sense, disregarding its objective, “consumer” effects. A “winner” on a congestion pricing

reform in a citizen sense is hence someone who approves of the underlying, abstract logic and rationale

of congestion pricing – whose views of fairness, environment and other societal dimensions are aligned

with what congestion pricing represents (for that individual). In order to estimate this “citizen utility”, a

model was estimated that separated how self-interest variables and attitude to various societal issues –

environment, taxes, pricing as an allocation instrument, equity and trust in government – affected

attitudes to congestion pricing. This allowed for an even broader definition of the “utility” of a

congestion pricing reform, capturing both consumer and citizen components. The distribution of these

“utility components” across income groups can then be explored.

This analysis showed that in terms of “citizen utility”, it is in fact middle-income groups who fare

better than average, with the exception of Helsinki, where groups fare better (relative to the average) the

lower their income is. Differences across income groups are relatively small, however.

In summary, it is hard to find much support of the view that congestion pricing is unfair, as long as

its purpose is to correct prices and allocate scarce resources. Both in terms of absolute payments and

compound self-interest, lower income groups fare better than average. From a citizen perspective,

differences are small, but lower income groups fare at least as well (and in some cases better) than high-

income groups. This changes, however, if the purpose of a charging system is in fact to generate

revenues. In that case, the regressivity of the pricing systems – that poor pay more relative to their

income – is a serious problem: it is difficult to defend that poor groups should contribute more than

proportionately to public revenues.



Notes

1. Other terms for essentially the same distinction are “homo economicus” vs. “homo politicus”, or

“personal well-being” vs. “subjective social welfare”.

2. It can be shown that the relative size of the adaptation cost to the tolls paid, i.e. the “triangle under the

demand curve” relative to the “rectangle under the demand curve” (ignoring changes in travel time for

the moment), given a relative price change a and demand elasticity ϵ is aϵ/2. To illustrate magnitudes,

assume that a congestion charge increases the monetary cost of an average car trip by 25%, and that the

demand elasticity is in the order of -0.5. This would give an adaptation cost which would be around 6%

of the toll paid. More careful quantitative analyses, separating welfare effects into time gains, adaptation

costs and paid tolls can be found in Eliasson (2009) and Eliasson and Levander (2006).

3. This is the average toll payment per income class divided by the average income in that class, which is

the resolution available in the data. An alternative measure would be the average of (toll payment divided

by income), but this causes problems for people with (notional) very low or even zero incomes.

4. This view is not shared by all, though. At the time I am writing this, a representative of the Swedish Left

party is quoted in a newspaper saying “Uniform pricing of trips is a necessary fairness reform.

Healthcare costs the same for everyone: travelling should, too.” However, it turns out that the view that

all trips should cost the same apparently only applies to public transport trips: the Left party is very much

in favour of congestion charges.

5. In Lyon, the hypothetical situation instead involved a closed tunnel, as this was judged to be closer to

reality and easier to imagine.

6. To check the robustness of the conclusions from this exploratory analysis, two other methods are also

used. First, a binary model is estimated, grouping the responses into positive or negative, discarding “no

opinion”. This reduces the problem with different distances between thresholds, since there will, in

essence, only be one “threshold”. The parameters of this binary model turn out to be close to the ordered

logit model, so conclusions stay unchanged. Second, the ordered logit model is used to predict the

responses of all the individuals, using only these self-interest variables. This will be a measure of how

individuals “should” vote if they only took self-interest into account. (Note that the reasons that this is

meaningful is that the model does not incorporate constants for each income group; in that case, the

model predictions had simply coincided with the actual average voting responses for each income class.)

There is also, as noted above, a risk that parameters are biased since attitudinal variables are omitted. To

check that, results are also compared with the model presented below where attitudinal variables are

included. It turns out that the parameters of the “self-interest” variables hardly change.

7. The word “fair” is, it turns out, not always easy to translate. The Swedish survey used the word rimlig

which can also be translated as “reasonable”, “acceptable”, “justified”.



References

Börjesson, M., Eliasson, J., & Hamilton, C. (2016). Why experience changes attitudes to congestion

pricing: The case of Gothenburg. Transportation Research Part A: Policy and Practice, 85, 1–16.

Börjesson, M., Eliasson, J., Hugosson, M. B., & Brundell-Freij, K. (2012). The Stockholm congestion

charges—5 years on. Effects, acceptability and lessons learnt. Transport Policy, 20, 1–12.

Börjesson, M., & Kristoffersson, I. (2015). The Gothenburg congestion charge. Effects, design and

politics. Transportation Research Part A: Policy and Practice, 75, 134–146.

Center for Public Policy Priorities. (2007). Who pays Texas taxes? ( No. No. 287). Center for Public

Policy Priorities.

Eliasson, J. (2008). Lessons from the Stockholm congestion charging trial. Transport Policy, 15(6), 395–

404.

Eliasson, J. (2009). A cost-benefit analysis of the Stockholm congestion charging system. Transportation

Research Part A: Policy and Practice, 43(4), 468–480.

Eliasson, J. (2014). The role of attitude structures, direct experience and framing for successful

congestion pricing. Transportation Research A, 67, 81–95.

Eliasson, J., & Jonsson, L. (2011). The unexpected “yes”: Explanatory factors behind the positive

attitudes to congestion charges in Stockholm. Transport Policy, 18(4), 636–647.

Eliasson, J., & Levander, A. (2006). Equity Effects of the Stockholm Trial. Transek report.

Eliasson, J., & Mattsson, L.-G. (2006). Equity effects of congestion pricing: Quantitative methodology

and a case study for Stockholm. Transportation Research Part A: Policy and Practice, 40(7), 602–

620.

Eliasson, J., Pyddoke, R., & Swärdh, J.-E. (2016). Distributional effects of taxes on car fuel, use,

ownership and purchases ( No. 2016:X). CTS Working Paper. Centre for Transport Studies, KTH

Royal Institute of Technology.

Gaunt, M., Rye, T., & Allen, S. (2007). Public Acceptability of Road User Charging: The Case of

Edinburgh and the 2005 Referendum. Transport Reviews, 27(1), 85–102.

Greene, W. H. (2003). Econometric analysis. Prentice Hall.

Hamilton, C. J., Eliasson, J., Brundell-Freij, K., Raux, C., & Souche, S. (2014). Determinants of

congestion pricing acceptability ( No. 2014:11). CTS Working Paper. Centre for Transport

Studies, KTH Royal Institute of Technology.

Hårsman, B., & Quigley, J. M. (2010). Political and public acceptability of congestion pricing: Ideology

and self-interest. Journal of Policy Analysis and Management, 29(4), 854–874.



Jaensirisak, S., May, A. D., & Wardman, M. (2003). Acceptability of Road User Charging: The

Influence of Selfish and Social Perspectives. J. Schade and B. Schlag (eds.): Acceptability of

Transport Pricing Strategies. Elsevier, Oxford.

Karlström, A., & Franklin, J. P. (2009). Behavioral adjustments and equity effects of congestion pricing:

Analysis of morning commutes during the Stockholm Trial. Transportation Research Part A:

Policy and Practice, 43(3), 283–296.

Le Vine, S., Jones, P., & Polak, J. (2013). The Contribution of Benefit-in-Kind Taxation Policy in Britain

to the “Peak Car” Phenomenon. Transport Reviews, 33(5), 526–547.

Levinson, D. (2010). Equity Effects of Road Pricing: A Review. Transport Reviews, 30(1), 33–57.

Metcalf, G. E. (1996). The lifetime incidence of state and local taxes: measuring changes during the

1980s. In J. Slemrod (Ed.), Tax progressivity and income inequality. Cambridge University Press.

Nyborg, K. (2000). Homo Economicus and Homo Politicus: interpretation and aggregation of

environmental values. Journal of Economic Behavior & Organization, 42(3), 305–322.

Sagoff, M. (1988). The economy of the earth. Cambridge: Cambridge University Press.

Schade, J., & Schlag, B. (2003). Acceptability of road user charging: meeting the challenge.

Acceptability of transport pricing strategies, 27.

Small, K. A. (1992). Using the revenues from congestion pricing. Transportation, 19, 359–381.

Suits, D. B. (1977). Measurement of Tax Progressivity. The American Economic Review, 67(4), 747–

752.

West, S. E. (2004). Distributional effects of alternative vehicle pollution control policies. Journal of

Public Economics, 88(3–4), 735–757.



Appendix: Estimation Results

Ordered logit model, self-interest variables only Value Std. Error t value toll -0.0010832 0.0001125 -9.6249 Toll, inc.grp 1 -0.0007957 0.0002209 -3.6030 No car 0.3061705 0.0610495 5.0151 cartrips 0.4306097 0.0284951 15.1117 Value of time, drivers 0.2389846 0.0189181 12.6326 Stockholm 0.9948645 0.0668150 14.8898 Helsinki 0.0213867 0.0740519 0.2888 Lyon -0.1766564 0.0722816 -2.4440 Gothenburg2014 0.5572110 0.0697637 7.9871 Intercepts: Value Std. Error t value 1|2 0.4265 0.1022 4.1721 2|3 1.4528 0.1034 14.0502 3|4 1.9839 0.1047 18.9533 4|5 3.3649 0.1098 30.6583 Residual Deviance: 19960.01 AIC: 19986.01 Binary logit model, self-interest variables only Estimate Std. Error z value Pr(>|z|) (Intercept) -1.9731613 0.1347027 -14.648 < 2e-16 *** toll -0.0009835 0.0001451 -6.778 1.22e-11 *** Toll, inc.grp 1 -0.0013063 0.0003366 -3.881 0.000104 *** No car 0.3865780 0.0772495 5.004 5.61e-07 *** cartrips 0.4892545 0.0365803 13.375 < 2e-16 *** Value of time, drivers 0.2467155 0.0239912 10.284 < 2e-16 *** Stockholm 1.2330181 0.0881791 13.983 < 2e-16 *** Helsinki 0.1313171 0.0977848 1.343 0.179298 Lyon -0.0158074 0.0914899 -0.173 0.862827 Gothenburg2014 0.6557825 0.0907974 7.222 5.10e-13 *** --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 (Dispersion parameter for binomial family taken to be 1) Null deviance: 8367.4 on 6078 degrees of freedom Residual deviance: 7229.9 on 6069 degrees of freedom (1011 observations deleted due to missingness) AIC: 7249.9 Ordered logit model, self-interest and attitude variables Value Std. Error t value Tolls -0.0010111 0.0001179 -8.5742 Tolls, add. inc.grp 1 -0.0005993 0.0002281 -2.6274 No car 0.3297529 0.0641394 5.1412 Car trips 0.3327746 0.0302396 11.0046 Value of time, drivers 0.1825060 0.0199794 9.1347 Stockholm 1.0063928 0.0713684 14.1014 Helsinki 0.1555103 0.0780314 1.9929 Lyon -0.0990411 0.0771785 -1.2833 Gothenburg2014 0.5615992 0.0776810 7.2296 environ 0.2609645 0.0166537 15.6701 TaxTooHigh -0.2524060 0.0118604 -21.2813 PricingIsFair 0.1172166 0.0114416 10.2447 AgencyIsFair 0.0504215 0.0098760 5.1054 equity -0.0042233 0.0134644 -0.3137 Intercepts: Value Std. Error t value 1|2 1.0315 0.1610 6.4059 2|3 2.1687 0.1627 13.3295



3|4 2.7289 0.1639 16.6526 4|5 4.2695 0.1684 25.3584 Residual Deviance: 17698.47 AIC: 17734.47 Binary logit model, self-interest and attitude variables Estimate Std. Error z value Pr(>|z|) (Intercept) -2.9264067 0.2270337 -12.890 < 2e-16 *** Tolls -0.0009104 0.0001588 -5.735 9.77e-09 *** Tolls, add. inc.grp 1 -0.0009653 0.0003484 -2.770 0.0056 ** No car 0.4413697 0.0867175 5.090 3.59e-07 *** Car trips 0.3767392 0.0406581 9.266 < 2e-16 *** Value of time, drivers 0.1845074 0.0268289 6.877 6.11e-12 *** Stockholm 1.3605294 0.1002557 13.571 < 2e-16 *** Helsinki 0.3196677 0.1084253 2.948 0.0032 ** Lyon 0.1205585 0.1029584 1.171 0.2416 Gothenburg2014 0.6459297 0.1068889 6.043 1.51e-09 *** environ 0.3026574 0.0230508 13.130 < 2e-16 *** TaxTooHigh -0.2899107 0.0160354 -18.079 < 2e-16 *** PricingIsFair 0.1538182 0.0158089 9.730 < 2e-16 *** AgencyIsFair 0.0664233 0.0135873 4.889 1.02e-06 *** equity -0.0224911 0.0179744 -1.251 0.2108 --- Signif. codes: 0 ‘***’ 0.001 ‘**’ 0.01 ‘*’ 0.05 ‘.’ 0.1 ‘ ’ 1 (Dispersion parameter for binomial family taken to be 1) Null deviance: 7943.2 on 5766 degrees of freedom Residual deviance: 6066.0 on 5752 degrees of freedom (1323 observations deleted due to missingness) AIC: 6096

International Transport Forum2 rue André Pascal F-75775 Paris Cedex [email protected]

Is Congestion Pricing Fair? - itf-oecd.org · Is congestion pricing fair? Consumer and citizen perspectives on equity effects Discussion Paper No. 2016-13 Prepared for the Roundtable

Documents